Stock portfolio optimization an‎d selec‎tion using multi-criteria decision making techniques in the Nasdaq stock market with a novel approach to forecast stock returns an‎d risk

9/14/2026 12:00:00 AM

Given the wide variety of stocks in financial markets an‎d the need for investors to selec‎t portfolios with maximum returns an‎d minimum risk, the aim of this research is to propose a comprehensive model for ranking, forecasting, an‎d optimizing stock portfolios in the NASDAQ market. This model seeks to integrate multi-criteria decision-making (MCDM) techniques, machine learning, an‎d multi-objective optimization to deliver superior performance compared to traditional approaches. In the first step, 100 NASDAQ stocks were ranked using 10 fundamental indicators an‎d six MCDM methods. By applying rank aggregation techniques, the top 13 stocks were selec‎ted. The returns of these stocks were then forecasted using Long Short-Term Memory (LSTM) neural networks across 24 separate models an‎d different time windows from 2000 to 2024. The results indicated that utilizing data from the most recent nine years led to higher prediction accuracy. Based on this, the returns of the selec‎ted stocks were predicted for October 1, 2024. Subsequently, sentiment analysis of financial news, the Relative Strength Index (RSI), an‎d fundamental scores were employed to extract three complementary indicators for each stock. Finally, an extended Markowitz model with three objective functions - risk minimization, return maximization, an‎d Shannon entropy maximization - was developed, incorporating constraints based on fundamental, technical, an‎d sentiment indicators. The model was solved through goal programming, an‎d the results showed that the obtained return (4.8% in one month) was nearly double the return of the NASDAQ index over the same period. Comparisons with traditional models, as well as tests involving the removal of certain components an‎d altering the return forecasting method, confirmed the high efficiency of the proposed model. This research, therefore, presents a realistic an‎d data-driven framework for investment decision-making.

بهينه‌سازي سبد سهام , تشكيل سبد سهام , نزدك , تصميم‌گيري چندمعياره , حافظه طولاني كوتاه‌مدت , تحليل بنيادي , تحليل احساسات

Portfolio Optimization , Portfolio selec‎tion , Nasdaq , MCDM , LSTM , Fundamental Analysis , Sentiment Analysis

Ali Beikvali

Dr. Emran Mohammadi